. Author manuscript; available in PMC: 2014 Jan 2.

Published in final edited form as: Stat Med. 2012 Mar 22;31(22):10.1002/sim.5351. doi: 10.1002/sim.5351

Table IV.

Simulation results comparing standard logistic regression, flexible logistic regression for pooling, and maximum likelihood approaches with regard to parameter and odds ratio estimates under different pooling conditions and sample sizes.

	Pooling approach			Estimates
Model	m pool size	N sample size	N/m assays	Mean ${\bar{θ}}_{1}$ ₁(SD)	Mean ${\bar{θ}}_{2}$ ₁ (SD)	Mean ln $(\bar{O R})$ (SD)	Mean $\bar{O R}$ (SD)
Logistic regression	1	1000	1000	0.208 (0.097)	0.502 (0.107)	0.656 (0.071)	1.932 (0.138)
Maximum likelihood	1	1000	1000	0.207 (0.087)	0.503 (0.096)	0.655 (0.071)	1.930 (0.137)

Logistic regression	1	2000	2000	0.198 (0.075)	0.503 (0.085)	0.647 (0.043)	1.912 (0.082)
Flexible logistic regression	2	2000	1000	0.199 (0.112)	0.505 (0.140)	0.649 (0.051)	1.916 (0.098)
Maximum likelihood	2	2000	1000	0.201 (0.084)	0.501 (0.107)	0.647 (0.050)	1.912 (0.096)

Notes: 1,000 replications; true values: θ1, = 0.20, θ2 = 0.50, ln(OR) = 0.646, OR = 1.908